Introduction to Pulsed Noise


Tutorial home




Before continuing, we’re going to make a little side trip and explore pulsed noise.  Pulsed noise is, simply, noise that’s gated by a VCA so that the noise appears in short pulses.


Pulsed noise has been detected in bowed string instruments.  The bow/string interaction creates more noise when the bow is slipping along the string than when the bow is sticking to it.  Pulsed noise has also been detected in reed woodwinds, created as a result of the reed hitting the mouthpiece.  Although it’s not so prevalent in flutes, we’re going to build a pulsed noise circuit and add it to our flute model, just to hear what it sounds like.


This particular technique of creating pulsed noise has been patented by Dr. Chris Chafe in US patent 5,157,216, and has been assigned to Stanford University.




Sparse Noise


The technique begins with the creation of sparse noise.  Sparse noise is white noise that has been modified to contain more “zero” outputs than would normally appear.  Below is a patch that creates sparse noise.





The density of the sparse noise is controlled by a single knob at the top of the patch called “Noise Probability”.  Let’s trace what it does:


  1. The top-most noise source creates white noise.  This is a random number between -64 and +64.  A new number is generated every sample.
  2. The noise goes through a full-wave rectifier.  This converts any negative number into its corresponding positive number (like a -10 into a +10), while leaving positive numbers unchanged.  The result is a signal that contains a string of random positive numbers.
  3. These positive numbers are compared with the value of the “Noise Probability” knob.  If the knob is higher than the noise value, the comparator’s output is logic high (+64).  If not, the comparator’s value is logic low (0).
  4. This logic signal is used to control a gate that switches a second noise source on or off.


So what happens when the “Noise Probability” is maximum (64)?  The knob is always >= the rectified noise source, so the gate is always on, and the final output signal is normal white noise, as seen below.



What happens when the “Noise Probability” is 0?  The knob is never greater than the noise source, so the gate is always off, and the final output signal is zero, as seen below.  (Actually, the gate will be on when the noise source is zero, but in that case the output will be zero anyway.)



And what happens when the “Noise Probability” is 8?  On average, the knob is >= the rectified noise source about 12.5% of the time.  This means that the gate is on about 12.5% of the time, and off about 87.5% of the time.  So the final output signal is white noise, filtered so that about 7/8 of the output samples are zero, as seen below.





Creating Pulses


The next step is creating pulses of the sparse noise.  Below is a patch that does this.





Here, the noise is being gated by the pulse width of an oscillator.  The pulse width is determined by the “Noise Threshold” knob.


So, the pulsed noise is controlled by two parameters: the probability of the sparse noise, and the width of the pulse.  Lowpass and highpass filters are included too, but are bypassed.  They’ll be turned on later, when we add this circuit to the flute model.


Below are two scope traces that display the results of various 50% and 25% pulse widths.  In both, the Probability knob is set to its maximum value, so that the noise is not “sparse”.




Below is a scope trace that displays a 25% pulse width, and sparse noise (also 25%).




Add pulsed noise to the flute


Let’s add this technique to the flute model, and hear what it sounds like.  We’ll keep the original noise source in the patch (but turn it off), so that we can easily switch back and forth between the two techniques.





The sparse noise is created as described above.  Pulses are then created by comparing a panel knob to the driver’s output level.  If the driver’s output level is greater than the knob’s value, the gate is turned on, and sparse noise is injected into the pipe.  Highpass and lowpass filters give us some basic tone control.


Five parameters control the noise, and are brought out to the front panel:


  1. Noise probability (the sparseness of the noise).
  2. Trigger threshold (pulse width).
  3. Highpass filter cutoff frequency.
  4. Lowpass filter cutoff frequency.
  5. Output level.


For good measure, the noise level of the “noisy air source” is also brought out to the panel, so the two methods can be compared.






Flutes don’t really have much pulsed noise anyway, so its inclusion in this model is just a matter of taste.


But pulsed noise will be especially useful in other models.  In woodwinds, it will help to stabilize the model and prevent the pipe from locking into higher modes.  In bowed strings, it will provide a more realistic bowed sound, especially on low notes.